Scaling observability: the real cost of Grafana, Datadog, and New

The multi-meter reality of Grafana Cloud

Grafana Cloud pricing in 2026 requires managing seven independent meters that bill for metrics, logs, traces, profiles, visualization, k6, and IRM. If you run a microservices stack that generates 100 GB of logs every day, you pay \$1,650 per month on the Pro tier because you pay \$0.05 per GB to process, \$0.40 per GB to write, and \$0.10 per GB for 30-day retention. This total includes \$5 per day for processing, \$1,350 per month for writes, and \$300 per month for 30-day retention. I notice that engineers often hit the 10,000 active series cap in the free tier quickly because a Kubernetes cluster with 50 pods sends enough telemetry to exceed that limit immediately. The free tier includes 50 GB of logs, 50 GB of traces, and 50 GB of profiles, but the 14-day retention limit prevents long-term trend analysis. The free tier also includes 2,232 host hours and 37,944 container hours per month. Metrics costs for the Pro tier sit at \$6.50 per 1,000 billable series for low resolution, though you pay \$16 per 1,000 series for high resolution. Grafana bills metrics based on the 95th percentile of active series to provide a buffer against usage spikes. Datadog avoids this specific complexity by using a per-host model, charging \$15 per host per month for Pro users on annual plans.

Comparing the enterprise pricing models

New Relic and Datadog target different budget priorities, but both create significant forecasting challenges. New Relic charges \$99 per month for full platform users, which means a 50-person engineering team pays \$4,950 per month in licensing alone before adding any data ingestion. You can also choose the Core user tier at \$49 per month or rely on the 100 GB free tier. New Relic data ingestion costs range from \$0.30 to \$0.50 per GB beyond that limit. Datadog scales based on the number of monitored hosts and the volume of logs ingested. I find the Datadog Pro tier at \$15 per host per month to be predictable, yet the cost for custom metrics or container restarts can spike your bill unexpectedly. For Enterprise users, Datadog charges \$23 per host per month on an annual basis.

You should look at the data volume before signing a contract. New Relic offers 100 GB of free data every month, but you pay \$0.30 to \$0.50 per GB for anything beyond that. Datadog provides 1,000-plus integrations, but its pricing varies by product and plan. Datadog also charges an additional cost per container beyond the agent monthly limit. I would skip the Datadog monthly billing if you cannot commit to an annual plan because the Pro tier jumps from \$15 to \$18 per host.

The high price of consolidated observability

Consolidating observability into a single tool often reveals that most ingested data stays unused. I see teams pay for the same 84% of unused data regardless of whether they choose Grafana, Datadog, or New Relic. While Grafana Cloud Pro offers a path for growing teams, self-hosting the open-source LGTM stack costs more in labor than the SaaS fees. Sirius estimates that internal SRE labor for self-hosting the full LGTM stack exceeds \$300,000 per full-time employee annually, a figure that dwarfs the cost of a managed service. Datadog revenue reached \$1.121 billion in Q2 2026, which proves the massive demand for these managed services. I find the New Relic Compute Capacity Unit model makes cost forecasting nearly impossible for many teams. New Relic provides 780-plus integrations, while Datadog offers over 1,000. Will the shift toward agentic coding change how much telemetry we need to store?